资源描述：
《DOA Estimation Using a Greedy Block Coordinate Descent Algorithm英文》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、6382IEEETRANSACTIONSONSIGNALPROCESSING,VOL.60,NO.12,DECEMBER2012DOAEstimationUsingaGreedyBlockCoordinateDescentAlgorithmXiaohanWei,YaboYuan,andQingLingAbstract—Thispaperpresentsanoveljointlysparsesignalutilizesthepropertythatthespatialspectraofthepointsourcesreconstructi
2、onalgorithmfortheDOAestimationproblem,aimingovertimearejointlysparse2.Hence[5]proposesan-normtoachievefasterconvergencerateandbetterestimationaccuracyminimizationformulationwhichpenalizesthejointsparsityofcomparedtoexisting-normminimizationapproaches.Theproposedgreedyblo
3、ckcoordinatedescent(GBCD)algorithmthespatialspectra;thisproblemisthensolvedinasecond-ordersharessimilaritywiththestandardblockcoordinatedescentconeprogramming(SOCP)framework.Recentworkalongthismethodfor-normminimization,butadoptsagreedyblocklineincludes[6],whereaweighted
4、-normminimizationfor-selectionrulewhichgivespreferencetosparsity.Althoughgreedy,mulationreplacesthestandard-normminimizationformu-theproposedalgorithmisprovedtoalsohaveglobalconvergencelationin[5].Anothernotableworkis[7],whichdealswithper-inthispaper.Throughtheoreticalan
5、alysiswedemonstrateitssta-bilityinthesensethatallnonzerosupportsfoundbytheproposedturbationsandinaccuraciesinthebasismatrixusingthebranch-algorithmaretheactualonesundercertainconditions.Last,weand-boundtechnique.However,theaboveapproachesareallmoveforwardtoproposeaweight
6、edformoftheblockselectionquitetime-consumingespeciallywhentheproblemdimensionrulebasedontheMUSICprior.Therefinementgreatlyimprovesislarge,e.g.,thesparsevectororthejointlysparsevectorsaretheestimationaccuracyespeciallywhentwopointsourcesarewithlargesize,thenumberofsnapshot
7、sislarge,etc.Toover-closelyspaced.NumericalexperimentsshowthattheproposedGBCDalgorithmhasseveralnotableadvantagesovertheexistingcomethiscomputationaldifficulty,[8]proposesacycliciter-DOAestimationmethods,suchasfastconvergencerate,accurateativemethodandsolvesasemi-definitep
8、rogramming(SDP)reconstruction,andnoiseresistance.problembyexploitingcovariance-baseddomainsparsity.Inde
